Jump to content

So I'm planning to build a system focussed on machine learning for home use since i can't always rely on resources of my university.

I'm located in Germany, but could also shop in Switzerland for parts if stuff's cheaper over there. My budget is set at 1500€.

I have a rough idea of what kind of specs I need, but considering my last PC build was a couple of years ago and I only loosely followed the innovations in this sector, I need help choosing components which go together well. The plethora of GPU-manufacturers and the "creative" naming schemes (to put it lightly) deter me from just doing it on my own. To put in simply terms: to casuals, I'm the tech guy, to proper hardware gurus who live and breathe that stuff (read: you) I'm not. Okay, enough of me, let's go.

 

Needs and Shoulds:

  • Needs to be able to handle huge datasets (it's machine learning after all) and the processing requirements which come with that kind of task
  • Needs to be fairly quiet, so adding an AIO for the GPU seems reasonable
  • Should be somewhat upgradeable, like adding a second GPU (that's the max though)
  • Should fit into a small form factor case. I'd love to put the machine into one of the slots of my Ikea Kallax shelf, which are 33x33x37cm

Does NOT need:

  • Any gaming prowess. The most sophisticated game will likely be minesweeper.
  • Any RBG-nonsense
  • Consideration for peripherals like mouse, keyboard or monitors. I have two 1080p screens for productivity, that should be enough for the start.

 

My idea of a system:

GPU:

A Turing-based GPU for fp16 performance. Considering my budget I'd love to get the RTX2080 (no, not the TI, sadly too expensive), the 2070 would suffice if there is absolutely no way to make it work with my budget.

 

CPU:

8core, I'm eying the i7-9700K. I'm not against AMD though, if that makes sense, works for my application and saves a bit of budget to spend more on GPU.

 

RAM:

16gigs, nothing flashy, memory clock is not as important as advertisers want to make us believe.

 

Storage:

1tb ssd, nvme would be wonderful if possible. 3tb hard disk for datasets.

 

Motherboard:

Honestly no idea, really depends on whether I choose team blue or team red after all.

 

PSU:

In my calculations, considering a possible second GPU (identical to the first), I shouldn't even get close to crossing the 1000W-border. Your recommendations on what I should pick?

 

Cooling:
An AIO cooler for the GPU because throttling there would really defeat the purpose of the build. also noise considerations. Does the CPU really need one though?

 

Case:

As stated in the needs and shoulds, a small one which still fits my parts. Room for a second GPU down the line would be really lovely, but if that simply isn't possible, so be it. Doesn't need to be flashy in any way shape or form, no windows needed, a simple black box would be awesome.

 

OS:

Windows for convenience, but that doesn't really play into my budget, I can simply pick up a license if I need one. I'll likely switch to Linux at some point in the future anyway.

 

 

So, that's it. It would be really lovely to hear your recommendations with actual products, since I trust your expertise more than my own

Link to comment
https://linustechtips.com/topic/1039256-machine-learning-build/
Share on other sites

Link to post
Share on other sites

PCPartPicker part list / Price breakdown by merchant

CPU: AMD - Ryzen 7 1700 3 GHz 8-Core Processor  (€178.04 @ Amazon Deutschland) 
CPU Cooler: be quiet! - Dark Rock Pro 4 50.5 CFM CPU Cooler  (€72.49 @ Amazon Deutschland) 
Motherboard: MSI - X470 GAMING PRO CARBON ATX AM4 Motherboard  (€179.90 @ Amazon Deutschland) 
Memory: GeIL - EVO SPEAR 16 GB (2 x 8 GB) DDR4-3000 Memory 
Storage: Crucial - MX500 500 GB 2.5" Solid State Drive  (€74.89 @ Mindfactory) 
Storage: Toshiba - X300 4 TB 3.5" 7200RPM Internal Hard Drive  (€93.55 @ Amazon Deutschland) 
Video Card: Gigabyte - GeForce RTX 2080 8 GB GAMING OC WHITE Video Card  (€539.00 @ Amazon Deutschland) 
Case: be quiet! - SILENT BASE 601 | SILVER ATX Mid Tower Case  (€101.94 @ Amazon Deutschland) 
Power Supply: SeaSonic - FOCUS Plus Gold 850 W 80+ Gold Certified Fully-Modular ATX Power Supply  (€118.84 @ Mindfactory) 
Total: €1358.65
Prices include shipping, taxes, and discounts when available
Generated by PCPartPicker 2019-02-28 20:39 CET+0100

 

 

here you go. pure business build. swap out the CPU when Zen 2 comes out.

 

850 watts is more than enough and the good aircooler will keep the build silent together with the silent series cases from bequiet

Link to comment
https://linustechtips.com/topic/1039256-machine-learning-build/#findComment-12348143
Share on other sites

Link to post
Share on other sites

Hey, thank you so very much! I had no idea about this partpicking website, that is awesome!

Unluckily, the prices stated from your build aren't displayed correctly when I actually went to the vendors stated, since there are sometimes hefty import fees associated, which brought your build up to somewhere around 1650€, mostly due to the GPU suddenly costing over 800€.

I then went and modified that list a bit to bring the cost back down and try to incorporate some of my shoulds:

 

PCPartPicker part list / Price breakdown by merchant

CPU: AMD - Ryzen 7 1700 3 GHz 8-Core Processor  (€173.89 @ Mindfactory) 
CPU Cooler: be quiet! - Dark Rock Pro 4 50.5 CFM CPU Cooler  (€72.49 @ Amazon Deutschland) 
Motherboard: ASRock - B450M PRO4 Micro ATX AM4 Motherboard  (€75.99 @ Amazon Deutschland) 
Memory: G.Skill - NT Series 16 GB (2 x 8 GB) DDR4-2400 Memory  (€86.90 @ Amazon Deutschland) 
Storage: Samsung - 970 Evo Plus 500 GB M.2-2280 Solid State Drive  (€123.53 @ Mindfactory) 
Storage: Toshiba - X300 4 TB 3.5" 7200RPM Internal Hard Drive  (€93.55 @ Amazon Deutschland) 
Video Card: Zotac - GeForce RTX 2080 8 GB Blower Video Card  (€687.99 @ Mindfactory) 
Case: Thermaltake - Core V21 MicroATX Mini Tower Case  (€61.90 @ Amazon Deutschland) 
Power Supply: SeaSonic - FOCUS Plus Gold 850 W 80+ Gold Certified Fully-Modular ATX Power Supply  (€118.84 @ Mindfactory) 
Total: €1495.08
Prices include shipping, taxes, and discounts when available
Generated by PCPartPicker 2019-02-28 23:10 CET+0100

 

A GPU AIO still isn't in, but oh well. The only promising one I managed to find was the Alphacool 240 GPX, and that is sold out at the moment anyway. I guess it's one step at a time. Furthermore, I changed the motherboard to a Micro ATX and the case to something equally small. As far as I can tell it should still fit though, and the dropped SLI support going from your X470 to my B450 chipset doesn't matter to me anyway. I had to change the RAM as well because it simply didn't show up anymore.

 

What's your opinion regarding those changes?

Link to comment
https://linustechtips.com/topic/1039256-machine-learning-build/#findComment-12348719
Share on other sites

Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now

×